Electrolyte tuning in dye-sensitized solar cells with N-heterocyclic carbene (NHC) iron(II) sensitizers

Introduction

The field of dye sensitized solar cells (DSCs) has developed significantly [1-3] since the pioneering publication of O’Regan and Grätzel [4]. Photoconversion efficiencies (η) of ≈11–14% have been realized with ruthenium-based [2], zinc(II) porphyrin-based [5-9] or metal-free organic dyes [10-12]. In a recent review [2], Nazeeruddin points to the fact that only incremental enhancements of the photoconversion efficiencies of ruthenium dyes have occurred during the last two decades, with N3 (first reported in 1993 [13]) and N719 [14] (Scheme 1) remaining the state-of-the-art dyes. The low natural abundance of ruthenium (≈0.001 ppm in the Earth's crust [15]) and its consequential high cost motivates thorough investigations of the use of dyes based on Earth-abundant metals. Among the first row d-block metals, iron, cobalt, nickel, copper and zinc have Earth-crustal abundances of ≈41,000, 20, 80, 50 and 75 ppm, respectively. We and others [16-19] have realized the potential of DSCs sensitized by bis(diimine)copper(I) complexes, and both we [20] and Dragonetti et al. [21] have demonstrated the viability of DSCs comprising both copper(I)-based dyes and copper(I)/copper(II) redox shuttles. Photoconversion efficiencies in the range of 3–5% have been achieved for DSCs containing copper(I) dyes [22-24]. However, based upon the 800-fold greater natural abundance of iron versus copper, the holy grail of sustainable DSCs is the use of iron-based sensitizers.

Ferrere and Gregg [25] were the first to report of the use of a simple iron(II) sensitizer in a functional DSC; in the presence of the co-adsorbant chenodeoxycholic acid (cheno, Scheme 2), a DSC containing cis-[Fe{bpy-4,4'-(CO₂H)₂}₂(CN)₂] (Scheme 2) achieved a short-circuit current density (J_SC) of 0.29 mA cm⁻² and an open-circuit voltage (V_OC) of 360 mV. However, little progress was made in this area [26] until the interest shifted to the use of iron(II) complexes incorporating N-heterocyclic carbene (NHC) ligands. Replacing polypyridyl ligands by NHCs leads to a remarkable increase in the lifetime of the ³MLCT excited state (1000× longer), considerably enhancing electron injection [27-33]. The homoleptic complex 1 (Scheme 2) is currently the best-performing iron(II) NHC sensitizer among a series of related complexes screened in n-type DSCs [32]. In MeCN solution, 1 exhibits an MLCT band at λ_max = 520 nm with an extinction coefficient, ε_max, of 16,200 dm³ mol⁻¹ cm⁻¹ [33]. Herein lies an advantage of NHC iron(II) complexes over bis(diimine)copper(I) dyes, since the latter absorb in the visible with values of ε_max of ≈5,000 dm³ mol⁻¹ cm⁻¹ [16]. Gros and co-workers have demonstrated that a DSC sensitized with 1 in the presence of the co-adsorbant cheno (Scheme 2) and with an I⁻/I₃⁻ redox couple in MeCN achieves values of J_SC = 0.41 mA cm⁻², V_OC = 457 mV, fill factor (ff) = 68% and η = 0.13%. These parameters compare with J_SC = 13.25 mA cm⁻², V_OC = 687 mV, ff = 67% and η = 6.1% for a reference DSC containing N719 [34]. Previous DSC investigations have focused on structurally modifying the NHC iron(II) complex while maintaining a common co-adsorbant and electrolyte [32,34]. It is well established for ruthenium(II), copper(I) and organic dyes that modification of the electrolyte can have a profound effect on DSC performance [17,35-39]. We now demonstrate the effects on the performance of DSCs sensitized by the NHC iron(II) dye 1 of varying the solvent or additives in an I⁻/I₃⁻-based electrolyte. We present data for fully masked DSCs to avoid overestimation of their performance [40].

Results and Discussion

Effects of solvent and ionic liquid

The working electrodes for the DSCs were prepared using commercial FTO/TiO₂ electrodes immersed in a MeCN solution containing a mixture of sensitizer 1 and the co-adsorbant cheno (see Experimental section). Duplicate cells were made for all measurements to ensure reproducibility of the data. An I⁻/I₃⁻ redox couple was used in all DSCs, and the electrolyte compositions for the first two series of investigations are given in Table 1. In typical I⁻/I₃⁻-based electrolytes, LiI and I₂ are present in a solvent such as MeCN, 3-methoxypropionitrile (MPN) or valeronitrile, with added ionic liquid, often 1-butyl-3-methylimidazolium iodide (BMII, Scheme 3). Additives are also present to tune the TiO₂ conduction band energy and reduce detrimental recombination of electrons and oxidized dye [41-44].

We initially focused on the effect of changing the electrolyte solvent from MeCN (present in the commercial Solaronix electrolyte AN-50 used by Gros and co-workers [34]) to the more viscous and higher boiling MPN. The choice of solvent is crucial for optimization of DSC performance and long-term stability [45,46]. Our starting point was the standard electrolyte (E1, Table 1) that we employ for DSCs with bis(diimine)copper(I) sensitizers. This electrolyte contains the additive N-methylbenzimidazole (MBI) which is known to enhance values of V_OC [47]. Electrolytes E1 and E1a differ only in a change in solvent from MPN to MeCN (Table 1), while on going from E1 to E2 (or E1a to E2a), the ionic liquid is changed from 1-butyl-3-methylimidazolium iodide (BMII) to 1,2-dimethyl-3-propylimidazolium iodide (DMPII, Scheme 3). The latter is also present in the commercial electrolyte AN-50. The performance parameters and J–V curves for DSCs sensitized with the NHC iron(II) dye 1 and containing electrolytes E1, E1a, E2 and E2a are shown in Table 2 and Figure 1. The data in Table 2 are referenced with respect to a DSC sensitized with N719 (Scheme 1) and in the right-hand column of Table 2, relative efficiencies are given with η(N719) set to 100%. Data for multiple cells confirm the reproducibility of the DSC performances (Table S1, Supporting Information File 1).

We first note that all DSCs exhibit good fill factors, indicating satisfactorily operating devices. The performances of the masked DSCs with electrolytes E1a and E2a (MeCN as solvent) are slightly lower than those reported by Gros [34], consistent with the masking [40] of the DSCs in the present study. Significantly, a change from MeCN to MPN enhances J_SC with little change in V_OC or the fill factor, and results in a gain in global efficiency. A further gain is observed upon changing the ionic liquid with values of J_SC in the order E4 > E2 > E1 ≈ E3 (Figure 2, Table 2 and Supporting Information File 1, Table S1). The use of EMIMPF in electrolyte E4 enhances both J_SC and V_OC. The improved performance observed with electrolyte E4 was verified for a set of four DSCs, with values of J_SC and V_OC in the ranges of 0.70–0.90 mA cm⁻² and 519–542 mV, respectively (Table S1, Supporting Information File 1).

Effects of additives

Each of electrolytes E1–E4 contains MBI (0.5 M), and we next investigated the electrolytes without this additive. Electrolytes E1b–E4b are compositionally analogous to E1–E4 but without MBI (Table 3). Comparison of the parameters in Table 2 and Table 4 demonstrates the effects of eliminating MBI. On-going from E1 to E1b, E2 to E2b, or E3 to E3b, an increase in J_SC coupled with a decrease in V_OC is observed resulting in an overall enhancement in η. For E2b, the efficiency relative to N719 of 8.5% (Table 4) is remarkable for an iron(II)-based sensitizer, and arises from a significantly enhanced value of J_SC. The result was verified using four DSCs (Table S3, Supporting Information File 1) for which values of J_SC in the range 2.31–2.58 mA cm⁻² and relative efficiencies of 7.8–9.5% were observed. However, the improvement in J_SC is at the expense of V_OC (292–374 mV for electrolyte E2b versus 426–427 mV for E2, Table S1 and Table S3, Supporting Information File 1).

The absence of the additive MBI from electrolyte E4b causes bleaching of the electrode when the dye-functionalized electrode surface comes in contact with the electrolyte; the original red colour of the dye-functionalized electrode becomes almost white. This has a dramatic effect on the value of V_OC, with a fall from 542 to 42 mV (Table 2 and Table 4). The decrease was verified using duplicate DSCs (Supporting Information File 1, Table S1 and Table S2). While we have no clear explanation of this phenomenon, we note with interest that Dyson and co-workers have observed extensive interactions between imidazolium cations and polyiodide anions in ionic liquids doped with I₂ [48] and this may be relevant to the use of ionic liquids in DSCs. The J–V curves in Figure 3 illustrate the effects on the performance of sensitizer 1 by using electrolytes with (E1–E4) and without (E1b–E4b) the additive MBI.

The promising performances of DSCs with dye 1 combined with electrolyte E2b encouraged us to tune the components further. In Table 3, electrolytes E2c–E2h and E4f–E4h are based on E2 and E4 with different concentrations of MBI or TBP. We first consider MBI. The effects of altering the electrolyte composition with respect to MBI are seen in the J–V curves in Figure 4 and in the DSC parameters in Table 4 and Supporting Information File 1, Tables S1–S3. The general trend in Figure 4 is for an increase in J_SC as the concentration of MBI decreases: E2 with 0.5 M MBI gives the lowest J_SC (0.54 mA cm⁻²) but this is enhanced to >2.3 mA cm⁻² by reducing the concentration of MBI to 0.01 M or removing the additive altogether. The data for multiple DSCs in Table S3 confirm the observation with values of J_SC in the range 2.31–2.58 mA cm⁻² for electrolyte E2b (no MBI) and 2.61–2.78 mA cm⁻² for electrolyte E2e (0.01 M MBI). DSCs with these two electrolytes exhibit values of V_OC in the range 292–374 mV (Supporting Information File 1, Table S3) and overall photoconversion efficiencies of 0.47–0.57% (or 7.8–9.5% with respect to N719 set to 100%). External quantum efficiency (EQE) curves are shown in Figure 5 and Figure S1 (Supporting Information File 1). The EQE spectra are broad with λ_max in the range of 480–520 nm, consistent with the incident photon to current efficiency (IPCE) λ_max of 500 nm reported by Gros [34]. The use of DMPII rather than EMIMPF in the electrolyte enhances the value of EQE_max from ≈6 to 15%. The highest EQE_max values of ≈12% for electrolyte E2b and ≈15% for E2e are consistent with the trends in J_SC discussed above. These values compare with EQE_max ≈ 2.3% for a DSC sensitized with 1 and using the commercial AN-50 electrolyte [34].

We now turn to the effects of using TBP as an additive, while retaining DMPII as the ionic liquid in the electrolyte. For ruthenium dyes such as N719 combined with an I⁻/I₃⁻ redox couple, it is well established that TBP leads to improved open-circuit voltage [44]. On the other hand, we have previously demonstrated that for a representative heteroleptic bis(diimine)copper(I) dye, the addition of TBP to a standard I⁻/I₃⁻-based electrolyte in MPN is detrimental to DSC performance [38]. In the current investigation, electrolytes E2f, E2g and E2h were prepared with different concentrations of TBP as additive (Table 3). The J–V curves shown in Figure 6 and the DSC parameters in Table 4 and Table S2 (Supporting Information File 1) demonstrate a significant decrease in J_SC even when TBP is present in 0.05 M concentration. A further increase in the concentration lowers J_SC further while the value of V_OC increases from 339 mV (no TBP) to 541 mV (0.5 M TBP). However, this gain is not sufficient to enhance the global efficiency which drops from 0.51% to 0.26% (Table 4). The trends are reproduced for duplicate DSCs (Table S2, Supporting Information File 1).

Electrochemical impedance spectroscopy (EIS)

Electrochemical impedance spectroscopy (EIS) is an important technique for the investigation of interfaces in DSCs [49,50]. Fitting of the Nyquist and Bode plots, which are used to describe the EIS results, leads to parameters including the recombination resistance (R_rec), electron/hole transport resistance (R_tr), charge-transfer resistance at the counter electrode (R_Pt) and the active layer surface chemical capacitance (C_µ). All experiments in the following discussion were performed at V_OC conditions. The equivalent circuit model used in this study consisted of five elements (Figure S2, Supporting Information File 1). A series resistance (R_s), a resistance (R_Pt) and a constant phase element (CPE1) to model a platinum counter electrode, an extended distributed element (DX1) which represented the TiO₂/electrolyte interface as a transmission line model, and a Warburg element (W_s) associated with diffusion of the electrolyte. The constant phase element was employed in this study because of the surface roughness [51,52].

We chose to focus on understanding the observations involving the MBI additive, and EIS studies were conducted for electrolytes E2b, E2c and E2e. Measurements and curve fitting were made for duplicate cells to confirm the trends discussed below; data for one cell for each electrolyte are presented. The key parameters of the EIS measurements are summarized in Table 5, and the Nyquist plots are shown in Figure 7. The series resistance (R_S, measured from zero to the start of the first semi-circle in the plot and arising from the charge resistance at the FTO/TiO₂ interface [53]) is constant (≈9 Ω) for the three DSCs (Figure 7b). The value of R_Pt is extracted from the first semi-circle in the Nyquist plot (Figure 7b) and Table 5 shows these values to be similar for all DSCs. In principle, two further semi-circles should be observed in the Nyquist plot, one corresponding to R_rec and one to the diffusion resistance of the charge carriers in the electrolyte (R_d). However, as seen in Figure 7, these may overlap. The most striking feature of Figure 7 is the far larger second semi-circle for the DSC with electrolyte E2c, corresponding to a very significant increase in R_rec (Table 5) when 0.1 M MBI was used. EIS measurements were also made for a DSC with electrolyte E2 containing 0.5 M MBI but we were unable to fit the data because of the extreme dominance of the second semicircle associated with an extremely high recombination resistance.

The electrolytes E2b and E2e with no MBI and 0.01 M MBI, respectively, (Table 3) have the same overall photoconversion efficiencies (η = 0.51%, Table 4) and similar values of J_SC (2.31 and 2.61 mA cm⁻²). In contrast, a ten-fold increase in the concentration of MBI to 0.1 M leads to significantly lower J_SC (0.69 mA cm⁻²) and poor DSC efficiency (η = 0.18%). Compared to electrolyte E2b with no additive, the presence of 0.01 M MBI in electrolyte E2e leads to a lower R_rec, a higher C_µ and a lower R_tr. A combination of higher capacitance and lower transport resistance is consistent with a higher value of J_SC (Table 5). Combinations of lower C_μ and higher R_rec for E2b and higher C_μ and lower R_rec for E2e lead to similar DSC performances.

Greater recombination resistance and low J_SC values are observed when fewer electrons are injected to the semiconductor. Comparable R_tr values, but extremely high R_rec values for E2c lead to low J_SC and, as a consequence, to a low photoconversion efficiency. The value of the electron lifetime (τ) and the diffusion length (L_d) are also important parameters for understanding the electron injection behaviour and these depend on the R_rec value [54]. A larger L_d and a longer lifetime as observed for electrolyte E2c result, respectively, in a higher electron density and lower charge loss in the semiconductor. For each electrolyte, the transport time (τ_t) is lower than τ, resulting in efficient electron transport through TiO₂. Trends in the electron lifetime, the diffusion length and the transport time are consistent with the V_OC values (Table 5) which increase in the order E2c > E2b > E2e.

DSC stability

For electrolytes E2b and E2e, DSC performances were measured over a 40 day period and demonstrated the stability of the devices. Figure S3 (Supporting Information File 1) shows the trend in overall efficiencies which were the result of a general gain in V_OC (Figure S4, Supporting Information File 1) compensating for the general decrease in J_SC over time (Figure S5, Supporting Information File 1).

Conclusion

We have shown that the performances of DSCs sensitized with the NHC iron(II) dye 1 originally reported by Gros and co-workers [34], can be significantly enhanced by tuning of the electrolyte composition, while retaining an I⁻/I₃⁻ redox shuttle. The use of MPN in place of MeCN as solvent in the electrolyte leads to an improvement in J_SC with little change in V_OC, resulting in enhanced photoconversion efficiency. The choice of the ionic liquid in the electrolyte is critical. Of the four ionic liquids investigated and with the additive MBI present at a concentration of 0.5 M, the order of DSC performance in terms of ionic liquid is EMIMPF > DMPII > BMII ≈ BMIMPF (right side of Figure 8). Removal of the MBI additive is highly detrimental to the DSC performance if the ionic liquid is EMIMPF with V_OC falling from 542 to 42 mV (Figure S6, Supporting Information File 1), but in contrast, leads to a significant improvement of J_SC values for DSCs in which the ionic liquid is DMPII, BMII or BMIMPF (Figure 8, left vs right sides). The addition of TBP is detrimental to J_SC (Figure 6 and Figure 8) while improving V_OC (Figure 6 and Figure S3, Supporting Information File 1). Further tuning of the electrolyte composition with MPN solvent and DMPII ionic liquid revealed that DSCs with 0.1 M MBI in the electrolyte performed slightly better than when MBI was absent. The best values of J_SC and V_OC were in the ranges of 2.31 to 2.78 mA cm⁻², and 292 to 374 mV, respectively, leading to photoconversion efficiencies in the range of 0.47 to 0.57%. These represent 7.8 to 9.3% relative to an N719 reference DSC set at 100%. Our investigation has shown clearly that tuning of the components of electrolytes with an I⁻/I₃⁻ redox shuttle is critical to progress in the field of NHC iron(II) sensitizers.

Experimental

General

Dye 1 was prepared as previously reported [31].

Solar cell fabrication

Commercial titania electrodes (opaque, Solaronix) were used for the working electrodes. Each was rinsed with EtOH and dried on a heating plate at 500 °C for 30 min. The electrodes were cooled to 60 °C and immersed in a MeCN solution (0.5 mM) of the iron(II) dye 1 containing (0.1 mM) chenodeoxycholic acid overnight. Each reference working electrode was made by dipping a commercial titania electrode in an EtOH solution (0.3 mM) of dye N719 (Solaronix) overnight. After soaking in the dye baths, the electrodes were removed, washed with the same solvent as used in the dye bath (in case of MeCN the electrodes were washed a second time with acetone) and dried with a heat gun.

Commercial counter electrodes from Solaronix (Test Cell Platinum Electrodes Drilled) were rinsed with EtOH and dried on a heating plate at 500 °C for 30 min. The TiO₂ electrodes and Pt counter-electrodes were combined using thermoplast hot-melt sealing foil (Solaronix, Test Cell Gaskets, made from Meltonix 1170-60 sealing film, 60 microns thick) by heating while pressing them together. The electrolyte was introduced into the DSC by vacuum backfilling through a hole drilled in the counter electrode and this was then sealed using hot-melt sealing foil and a cover glass.

The solar cell measurements used fully masked cells using a black coloured copper sheet with a single aperture placed over the screen printed dye-sensitized TiO₂ square. The area of the aperture in the mask was smaller than the active area of the dye-sensitized TiO₂ (0.36 cm²). For complete masking, tape was also applied over the edges and rear of the cell. Current density–voltage (J–V) measurements were made by irradiating from behind with a LOT Quantum Design LS0811 instrument (100 mW cm⁻² = 1 sun at AM 1.5) and the simulated light power was calibrated with a silicon reference cell.

Electrolyte preparation

For DSCs with dye 1, electrolyte compositions are given in Table 2 and Table 4. BMII and TBP were purchased from Sigma-Aldrich, and DMPII, BMIMPF, EMIMPF, MBI, MPN were bought from TCI, Apollo Scientific, Fluorochem, Alfa Aesar and Fluka, respectively. MeCN was HPLC grade. For DSCs with N719 dye, the electrolyte comprised electrolyte E1.

EIS measurements

A ModuLab^® XM PhotoEchem photoelectrochemical measurement system from Solartron Analytical was used for the EIS measurements. The impedance was measured at the open-circuit potential of the cell at a light intensity of 22 mW cm⁻² (590 nm) in the frequency range of 0.05 Hz to 100 kHz using an amplitude of 10 mV. The impedance data were analysed using the ZView^® software from Scribner Associates Inc.